Operating Automations

Test an automation in Draft, prepare it for Live operation, then monitor and recover its runs.

At a glanceVerified in app · Oct 30, 2026

Use it to: Switch an automation between Draft and Live, set concurrency rules, and configure failure emails
Find it: Top canvas status bar, Workflow Settings → Concurrency, and Team settings → General
Limits: Max concurrent runs: 1 to 50 (up to 1,000 queued runs) · Keyed lock up to 256 characters and 1,000 queued per key

The operating loop

1. TestVerify in DraftTest isolated Steps and run end to end from the trigger using safe destinations.2. Go LiveSave a version and activateSnapshot the working state, complete the launch checklist, and switch triggers to Live.3. MonitorTrack workspace activityFollow run volume, pending approvals, and failed runs in Activity and the Execution Log.4. DiagnoseInspect Step payloadsFind the first failing Step in the Execution Log and inspect its inputs and outputs.5. RecoverRetry or route failuresConfirm whether external writes happened before retrying, and add error paths for resilience.

1. Test the path in Draft

Draft mode pauses automatic triggers while allowing manual runs. Use test accounts and safe destinations: manual runs still execute real external writes and spend credits. See Testing & Debugging.

Save a version before a risky change

Save a named restore point in Versioning so you can roll back at any time.

Verify uncertain Steps

Run with test data only replays saved sample output. Use an authorized Workflow Assistant test for app actions, or Test & Debug (Preview / Live test) for AI Steps.

Run a quick test from the trigger

Set the trigger as the run point and click Run (or Run a quick test) with realistic payloads to test the full path.

Check the values that leave the automation

Inspect the destination app to confirm that the created record, message, or file mapped as expected.

Before you turn it on

These checks are for your review only. They are not saved, and Glow does not verify or change your automation.

0 of 6 checked

Accounts and destinations
Verify connected accounts, recipients, and target records. Replace test addresses and retest changed destinations. See Managing Connections.
Test coverage
Run from the trigger across each branch, including missing-field cases, and check the destination app. Follow Testing & Debugging.
Failure handling
Configure retries and error paths. Check whether external writes already succeeded before repeating them. See Error Handling & Retries.
Usage and capacity
Estimate daily Step credits, Glow AI tokens, and storage in Billing & Usage, accounting for loops, retries, and per-Step rates.
Trigger setup
Confirm Scheduler timezones and Next runs, or send one controlled Webhook request before enabling production traffic.
Someone responsible
Assign an owner and backup with access to Activity to monitor initial runs and maintain connections.

Draft is not a sandbox. Manual runs can send real messages, change records, and use your allowance. A failed run does not prove that no external change occurred.

2. Set the automation Live

Safe editing & manual testingDraft Status

Pauses automatic schedules, polling triggers, and Webhook endpoints.
Does not queue missed trigger events while switched off.
Switch to Draft and save a version before making structural edits.

Production traffic armedLive Status

Validates trigger setup and enabled Wait Steps upon activation.
Arms Webhook URLs, polling schedules, and Scheduler timezones.
Edits apply immediately to new runs; inspect the first live execution.

Concurrency & queueing

By default, incoming trigger and schedule events start immediately in parallel. To cap overlapping runs or serialize events per record, open More (⋮) → Settings → Concurrency & queueing (concurrencyMode):

Concurrency modes

Mode	Value	Behavior
Unlimited (default)	`"unlimited"`	Every trigger or schedule event starts immediately with no per-automation cap.
Limit concurrent runs	`"max_concurrent"`	Runs up to Runs at a time (`1` to `50`). Set When the limit is reached to Wait in line (first in, first out) (`"queue"`, up to `1,000` queued) or Skip the new run (`"skip"`).
One at a time (skip overlapping)	`"singleton_skip"`	Runs at most one execution at a time and skips new trigger or schedule events while active.
Serialize by key	`"keyed_lock"`	Extracts a key from the Trigger step and Field (e.g. `{{ 1.body.customerId }}`, up to `256` characters; event triggers only, not Scheduler). Same-key runs execute in order (up to `1,000` queued per key); distinct keys run in parallel.

3. Monitor runs and workspace usage

Workspace Dashboard & Activity

Dashboard Overview: Track run volume, success rates, and allowance usage across 24-hour, 7-day, or 30-day windows.
Activity: Filter the daily workspace event stream by Runs, Automations, People, or Connections.
Billing & Usage: Review credit and token burn rates against your monthly plan.

4. Diagnose and recover

Use Troubleshooting & Common Errors when a run fails to start, stays waiting, or returns an error. Check whether external writes took effect before re-running a failed execution.

For automated alerting and recovery:

Configure retries and fallback paths in Error Handling & Retries.
Enable Failure emails in Team settings → Notifications (up to 3 recipients, grouped hourly).
Build a dedicated alert workflow with the Error Trigger.

Verify it works

Run: Save a version in Documentation & history, run a test from the trigger in Draft, and switch the status control to Live
You should see: Automatic triggers arm on the automation, and new runs appear in the Execution Log and workspace Activity.

What’s Next?

👉 Testing & Debugging →: test individual steps and run quick tests from the canvas.
Activity: monitor automation executions and workspace alerts.
Versioning: manage saved versions and understand Draft versus Live operation.